OCR is a well-known technique in recognizing either hand-written characters or scanned characters. With the recently development of the network, including Internet and intranet, the remote OCR system that need not each user to have an individual OCR apparatus has been expanded fast and broadly.
However, most of the existing remote OCR systems are running in a very simple way, just consisting of uploading (the image file for OCR) and downloading (the recognition result). User can not get control of the recognition process flow. For example, user can not select the image areas to be recognized, user can not use the candidate characters generated in recognition for manual proof.
Besides the above mentioned OCR functions, the efficiencies of the OCR systems also should be considered. In the traditional way, data of the original image file must be transmitted along with other data received from OCR operations. Accordingly, the more operations it takes, the more frequent transmissions are required. Due to effects of the server's hardware limit and the network bandwidth, it will eventually cause the inefficiency of the whole system.